Mapeamento digital de areia, argila e carbono orgânico por modelos Random Forest sob diferentes resoluções espaciais

Bhering, S. B.; Chagas, C. da S.; Carvalho, Waldir de; Pereira, N. R.; Filho, B. Calderano; Pinheiro, Helena Saraiva Koenow

doi:10.1590/s0100-204x2016000900035

Cited by 11 publications

(8 citation statements)

References 26 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…These results corroborate the hypothesis that the topography has an influence, in a larger spatial context, and has prevalence on prediction of soil particle size contents in the tested basin. In contrast, a case study with Random Forest with 30 m and 90 m DEM did not achieve significant differences in prediction [61]. Despite some similarity with covariables importance, like Elev and RSP, the modeling is done on single scale datasets.…”

Section: Prediction Of Superficial Layer Soil Texturementioning

confidence: 89%

Multiscalar Geomorphometric Generalization for Soil-Landscape Modeling by Random Forest: A Case Study in the Eastern Amazon

Araújo¹,

Oliveira²,

Beldini³

2021

JGIS

View full text Add to dashboard Cite

Multiscalar topography influence on soil distribution has a complex pattern that is related to overlay of pedological processes which occurred at different times, and these driving forces are correlated with many geomorphologic scales. In this sense, the present study tested the hypothesis whether multiscale geomorphometric generalized covariables can improve pedometric modeling. To achieve this goal, this case study applied the Random Forest algorithm to a multiscale geomorphometric database to predict soil surface attributes. The study area is in phanerozoic sedimentary basins, in the Alter do Chão geological formation, Eastern Amazon, Brazil. The multiscale geomorphometric generalization was applied at general and specific geomorphometric covariables, producing groups for each scale combination. The modeling was run using Random Forest for A-horizon thickness, pH, silt and sand content. For model evaluation, visual analysis of digital maps, metrics of forest structures and effect of variables on prediction were used. For evaluation of soil textural classifications, the confusion matrix with a Kappa index, and the user's and producer's accuracies were employed. The geomorphometry generalization tends to smooth curvatures and produces identifiable geomorphic representations at sub-watershed and watershed levels. The forest structures and effect of variables on prediction are in agreement with pedological knowledge. The multiscale geomorphometric generalized covariables improved accuracy metrics of soil surface texture classification, with the Kappa Index going from 43% to 62%. Therefore, it can be argued that topography influences soil distribution at combined coarser spatial scales and is able to predict soil particle size contents in the studied watershed. Future development of the multiscale geomorphometric generalization framework could include generalization methods concerning preservation of features,

show abstract

Section: Prediction Of Superficial Layer Soil Texturementioning

confidence: 89%

Multiscalar Geomorphometric Generalization for Soil-Landscape Modeling by Random Forest: A Case Study in the Eastern Amazon

Araújo¹,

Oliveira²,

Beldini³

2021

JGIS

View full text Add to dashboard Cite

show abstract

“…Thirty-one environmental variables (Table 2) were derived from an acquired 10 m spatial resolution digital elevation model (DEM) to create the dataset using SagaGIS software (Conrad et al, 2015). The covariates derived from the DEM are those most commonly used by DSM users (Calderano Filho et al, 2014;Carvalho Junior et al, 2014;Bhering et al, 2016;Chagas et al, 2016;Camera et al, 2017;Gruber et al, 2017;Heung et al, 2017).…”

Section: Environmental Covariatesmentioning

confidence: 99%

Sample design effects on soil unit prediction with machine: randomness, uncertainty, and majority map

Carvalho¹,

Pereira²,

Filho³

et al. 2020

Revista Brasileira De Ciência Do Solo

Self Cite

View full text Add to dashboard Cite

Notwithstanding the importance of soil surveys, advances in digital soil mapping have mainly focused on mapping soil attributes or properties rather than developing digital maps of soil units or soil classes. The purpose of this research was to develop digital soil unit maps based on primary soil data collection in areas without previously collected soil information. The covariate variability, the random effect across the data subset and the map outputs were the focuses of this study. We used five datasets with four models (Random Forest -RF, Gradient Boosted Machine -GBM, C5.0, and multinomial log-linear model -MLR). The covariates were grouped into five datasets, where four were grouped by Region Of Interest per Class (ROIC) and one was not grouped by ROIC. To evaluate the random effect to split the dataset, we ran each model 50 times and observed the overall accuracy (OA) and kappa index, and uncertainty, majority and variety maps. The OA of Dataset01 to 04 was lower than to Dataset05 accuracy. However, map outputs of RF and GBM for Dataset01 and Dataset05 had the same majority prediction. It seems that RF and GBM produce consistent results in map outputs according to this methodology and pedologist expertise. To evaluate the uncertainty and the consistency of soil unit prediction, we used the majority maps process. Random Forest, similar to GBM, presented the best results. The increase in the number of covariates was not a guarantee of improvement in the OA or in the quality of the map output. Geographic position and distance raster did not improve the map output according to expert evaluation. Because the variance between the ROICs, when the training and validation datasets were split based on it, the subsets are quite different in relation to the covariates, and this is the reason for the worse results of this model, comparing with the Dataset05. On the other hand, when considering one complete dataset not based on ROICs, the variance of training and validation subsets is lower and produced more accurate parameters of quality.

show abstract

“…O software utilizado neste artigo para o treinamento e aplicação dos algoritmos foi o WEKA -Waika to Environment for Knowledge Analysis (GARNER, 1995). Neste estudo optou-se por testar Árvore de Decisão (QUINLAN, 1993), devido à sua simplicidade e legibilidade; random forest (BREIMAN, 2001), pelos bons resultados apresentados em outras pesquisas (INZA et al, 2009;BHERING et al, 2016); e Redes Neurais artificiais pela robustez e habilidade em lidar com dados muito complexos.…”

Section: Variáveis Ambientaisunclassified

Algoritmos De Aprendizagem De Máquina Na Modelagem Da Distribuição Potencial De Habitats De Espécies Arbóreas

Carvalho

França

Lopes

et al. 2019

NAT

View full text Add to dashboard Cite

O estudo teve como objetivo avaliar três métodos de aprendizagem de máquina (árvore de decisão-J48, random forest e redes neurais artificias), na modelagem da distribuição de dez espécies arbóreas mais abundantes em uma sub-bacia do rio São Francisco (MG). Utilizaram-se dados provenientes do Inventário Florestal de Minas, com total de 77 fragmentos amostrados e 2.234 parcelas, nas quais foram computadas a presença/ausência de cada espécie. Empregaram-se 12 variáveis ambientais categóricas procedentes do Zoneamento Ecológico Econômico de Minas Gerais (ZEE/MG), além de variáveis relacionadas ao balanço hídrico do solo (evapotranspiração atual e potencial, aridez e índice alpha). A parametrização dos três algoritmos para as dez espécies selecionadas foi feita com o auxílio do algoritmo cv parameter do software WEKA. Os resultados mostram que os algoritmos testados apresentaram desempenhos estatisticamente iguais em 60% das espécies arbóreas. Os algoritmos random forest e multilayer perceptron foram estatisticamente iguais para a espécie Eugenia dysenterica, sendo superiores ao algoritmo J48. Contudo, o algoritmo random forest foi superior aos demais para as três espécies do gênero Qualea. Conclui-se que o algoritmo random forest apresentou-se como o mais robusto para a modelagem da distribuição potencial de habitat de espécies arbóreas.Palavras-chave: inteligência artificial; árvore de decisão; random forest; redes neurais artificiais. MACHINE LEARNING ALGORITHMS FOR MODELING THE POTENTIAL DISTRIBUTION HABITAT OF TREE SPECIES ABSTRACT: The aim of the present study was to evaluate three methods of machine learning (decision tree-J48, random forest and artificial neural networks) to model the potential habitat distribution of the ten most abundant tree species of the São Francisco river watershed. The presence/absence tree species data were from 77 fragments sampled with 2,234 plots. We used 12 categorical environmental variables from the Economic Ecological Zoning of Minas Gerais (ZEE/MG), as well as variables related to soil water balance (current and potential evapotranspiration, aridity and alpha index). The parameterization of the three algorithms was done with cv parameter algorithm of the WEKA software. The results showed the applied algorithms were statistically similar for 60% of the tree species. The random forest and multilayer perceptron algorithms were statistically similar considering the Eugenia dysenterica and superior to J48 algorithm. However, the random forest algorithm was superior to the other for the three species of Qualea genera. The conclusion is the random forest was the most robust model for the potential distribution habitat of tree species.Keywords: artificial intelligence; decision trees; random forest; artificial neural networks.

show abstract

Mapeamento digital de areia, argila e carbono orgânico por modelos Random Forest sob diferentes resoluções espaciais

Cited by 11 publications

References 26 publications

Multiscalar Geomorphometric Generalization for Soil-Landscape Modeling by Random Forest: A Case Study in the Eastern Amazon

Multiscalar Geomorphometric Generalization for Soil-Landscape Modeling by Random Forest: A Case Study in the Eastern Amazon

Sample design effects on soil unit prediction with machine: randomness, uncertainty, and majority map

Algoritmos De Aprendizagem De Máquina Na Modelagem Da Distribuição Potencial De Habitats De Espécies Arbóreas

Contact Info

Product

Resources

About